W3C

Publishing Maintenance Working Group Telco

22 January 2026

Attendees

Present
Abe Jellinek, Avneesh Singh, Brady Duga, Charles LaPierre, Dale Rogers, Grigorily Manucharian, Gregorio Pellegrino, Hadrien Gardeur, Ivan Herman, Laurent Le Meur, Masakazu Kitahara, Matt Garrish, Romain Deltour, Shinya Takami, Susan Neuhaus, Toshiaki Koike, Wendy Reid
Regrets
-
Chair
Susan Neuhaus, Wendy Reid
Scribe
Brady Duga, Wendy Reid

Meeting minutes

Wendy Reid: Please associate github with w3c accounts
… you can do it at github, so we can tag you

JXL in EPUB - w3c/epub-specs#2896

Wendy Reid: JXL is a new JPEG thing

Ivan Herman: Yes, a new JPEG, it's more compact, etc. I am not an expert
… It is an ISO standard, so we can refer to it
… Chrome has it now behind a flag
… firefox too. So it will be in major browsers soon
… authoring tools should be able to support it within the year
… so the question is, should we make it a core type?
… this brings us to around 7 image formats
… there was a long discussion around implementations
… there are some
… Shinya Takami brings up that reading systems must understand that media type
… if true that may be a problem for some reading systems
… but they also don't support webp which is also in the list

Avneesh Singh: Recalling opus, etc, the decision was to put it into the spec. If at the last minute we find there are no implementations we can pull it
… this gives a warning that it is coming even if it isn't in the spec release

Ivan Herman: At this moment the tests pass at least 2 implementations
… which is the required number

Avneesh Singh: I was talking broader industry support

Wendy Reid: We haven't mentioned webkit/safari
… Some ereaders use webkit
… we ran into this with webp
… should we have the list of well supported media types

<Susan Neuhaus> the can i use for fxl on the web https://caniuse.com/?search=jxl

Wendy Reid: should we have something like emerging media types? These would be flagged as not yet fully supported

Hadrien Gardeur: JXL has an interesting feature set for images
… it makes sense to have it
… there has been a battle over this, between AVIF and jxl
… we have core media for images, but not audio or video
… I am not really sure how whitelisting even helps
… the big question is should we keep track?

Matt Garrish: Core media type mainly means you don't need a fallback
… so technically it is a new requirement to support it
… to the other question, core media types made sense for 3.0
… but once you start adding to it, it doesn't make sense
… So you either have to pick between broad support and new features
… so I think media types doesn't make a lot of sense for general resources, just for spine items
… even if it is emerging, it will never be supported in old reading systems

Dale Rogers: The issues seem to be around media types, core media types and support types
… so if we are trying to align with the web it makes sense
… as an author if JXL has the features we want, shouldn't we just use it with a fallback?

Laurent Le Meur: we has this discussion with webp
… people using webp and jxl don't want to add a fallback if they know it is supported
… so we have core and extended core and it is a mess of our making

<Ivan Herman> +1 to Laurent Le Meur

Laurent Le Meur: so maybe we have a note that these are added after 3.0, so if you are worried you should add a fallback

Shinya Takami: if we can do a better solution, 2 lists
… one is RS MUST support, no fallback needed
… and one is RS SHOULD support and fallbacks SHOULD be used
… webp is not broadly supported in Japan
… jxl is still pretty new to add

Susan Neuhaus: would we make jxl a foreign resource?
… which requires a fallback

Brady Duga: One point, the concept of core media types goes back to the start
… early days we had discussions on whether to use GIF or JPEG, even though browsers didn't support, but we got lucky and the world went that way
… it does feel a bit ridiculous to keep adding to this list

<Susan Neuhaus> +1 Brady Duga

Brady Duga: this is guidance to authors, if we keep adding things that aren't fully supported, it's not helping authors
… webp was probably a mistake. We should be aiming for broad support
… the list of core media types was to give confidence, and if we want to allow people to use new media types without fallbacks, we could alternately just allow that
… people know they are using the wrong type of image, but they aren't forbidden
… putting a fallback in for webp or jxl is silly because the point is size
… we need a different solution that just adding to core media types

<Wendy Reid> +1 brady

Hadrien Gardeur: I want to go back to how this works on the web
… on the web you store the formats, and something else or generate it on the fly
… the tendency on the web is to go for the latest and greatest and provide the older format when needed
… epub isn't like that, we can't generate it
… but also we can't never advance because of old formats
… I disagree with Shinya Takami - multiple reading systems support avif and webp
… they just convert
… manga with jpeg is much bigger than the newer formats

Matt Garrish: in terms of authoring, a note makes more sense. Warnings are a problem for some authors and will require fallbacks anyway
… how bad is it that old reading systems don't support this?

<Ivan Herman> +1 to Matt Garrish

Grigorily Manucharian: The usage is important. webp is good, but usage is only 12%
… so despite this, people still prefer what they know

<Hadrien Gardeur> +1 for having a note about support instead of core media + fallback

Grigorily Manucharian: we need to understand where this is used

Wendy Reid: I agree that the current state is a problem
… I understand the a reading systems can convert to another format, but that is more likely to work within an ecosystem
… not when I just open a downloaded epub
… And we still have to consider offline
… we don't want to inundate people with warnings, but people look to us for a set of rules
… so if we don't mention JXL people will assume you can't use it
… I think we may have jumped the gun with webp
… so we need to say there are other formats, but they might be under implemented
… so we need to say "here are the known to work" types, and here are some that might not be as well implemented

Shinya Takami: SOme stores in Japan use webp
… it is very complicated in Japan
… having a format in core that publishers can't use isn't a good situation

Dale Rogers: I am hearing some formats are more or less popular
… as an author, I will use what works
… but something has to move the spec forward
… what eventually gets the publishers attention to support a particular format?
… what gets the reading systems attention to decide to implement?

Wendy Reid: We don't seem to be converging, so we need to move to the next one soon

Ivan Herman: if we do nothing, and they want the quality of JXL ,then they need fallbacks
… if they use without fallbacks epubcheck complains
… I would prefer to have it in the document
… knowing that it may not work
… my favorite option is to put in a note to say these formats are new and may not work everywhere
… this is not a unique thing in the spec

Matt Garrish: I agree, note is best. If we want to allow use without fallback we need to add to core
… I don't want to confuse things by adding a new type
… maybe also make foreign types not so scary

Selectors for Annotations - w3c/epub-specs#2892

Laurent Le Meur: [sharing screen]
… in an annotation there is a target
… the src references the epub
… the file in the epub
… the selector just deals with the exact points in the book that are part of the annotation
… [walking through example]
… in an annotation there can be multiple selectors, but they must point at the same place
… our first selector is text fragment
… this is from a community group, not a proper standard
… we will need to make TextFragmentSelector, as it isn't in webannotations
… the format is easy for that selector, it can be made a little trickier
… some notes in the annotation doc
… 1. we don't need to percent encode since it isn't in a url
… if there is a constrained on copied characters, you should use it with care
… don't put too much text!
… in the original model there was a text quote selector
… we can consider this new one as a modernized version of that selector
… one negative is there is no API currently in scripting, and no good polyfills
… so you have to manually walk the DOM
… also 1 question - should we add url fragment syntax? I didn't as I wanted to keep it short, and the static text doesn't seem necessary
… so should we add the static text

<Susan Neuhaus> ack: Ivan Herman

Ivan Herman: It is a CG document, but it is also the topic of active work in HTML
… I understand the argument with encoding
… but isn't this hiding the underlying URL fragment?
… since that is how it will be done in HTML
… that avoids parallel specing

<Hadrien Gardeur> +1 for what Ivan Herman just said, we should use FragmentSelector + conformsTo instead of creating TextFragmentSelector

Ivan Herman: and the last thing is, if I select a text, and some is in an html element, how is that handled

Laurent Le Meur: You just keep the text and ignore the elemenrs

<Susan Neuhaus> ack: Brady Duga

<Susan Neuhaus> Brady Duga: How well is the processing model for this group developed?

<Susan Neuhaus> …is the detail about how to do this specified somewhere?

<Susan Neuhaus> @Ivan Herman: it is being considered in the HTML group

<Hadrien Gardeur> Old document about this: https://wicg.github.io/scroll-to-text-fragment/

<Susan Neuhaus> Brady Duga: then it is scary to me to put this in our spec

Minutes manually created (not a transcript), formatted by scribe.perl version 244 (Thu Feb 27 01:23:09 2025 UTC).